2016 SOUTH AFRICAN RESEARCH CHAIR IN SOCIAL CHANGE, CENTRE FOR SOCIAL CHANGE, UNIVERSITY OF JOHANNESBURG. 2

Transcription

1 1

2 2016 SOUTH AFRICAN RESEARCH CHAIR IN SOCIAL CHANGE, CENTRE FOR SOCIAL CHANGE, UNIVERSITY OF JOHANNESBURG. 2

3 South African Research Chair in Social Change Report #3 LOCAL GOVERNMENT ELECTIONS 2016: SOME PRELIMINARY FINDINGS FROM AN EXIT POLL OF VOTERS Marcel Paret Centre for Social Change, University of Johannesburg 3

4 Acknowledgments This report has been produced under the auspices of the South African Research Chair in Social Change, which is funded by the Department of Science and Technology, administered by the National Research Foundation, and hosted by the University of Johannesburg (UJ). We are obliged to all these institutions, without which our research would not be possible. The construction of the survey and the coordination of the fieldwork drew upon a wide range of people within the Centre for Social Change. Peter Alexander, Heidi Brooks, Boituemlo Maruping, Boikanyo Moloto, Trevor Ngwane, Carin Runciman and Luke Sinwell assisted in the construction of the questionnaire and provided assistance throughout the process. Recruitment of fieldworkers and logistics was coordinated by Lucinda Becorny, Boitumelo Maruping and Boikanyo Moloto. Additional assistance was provided by Cassidy Landen, Craig Miller, Yoemna Saint and Thandazile Zwane Team leaders managed the survey on the day, often encountering and overcoming a number of difficulties. The team leaders were: Peter Alexander, Heidi Brooks, Hibist Kassa, Eunice Khumalo, Pinky Manong, Boitumelo Maruping, Thembelihle Maseko, William Mkhuma, Lebogang Mohlue, Kgothatso Mokgele, Boikanyo Moloto, Lesley Moyo, Paddington Mutekwe Bridget Ndibongo, Kamogelo Nthite, Marcel Paret, Mahlatse Rampedi, Carin Runciman, Siyabonga Sibiya, Fikiswa Simelane and Luke Sinwell. The survey could not have been undertaken without the hard work of all of the fieldworkers. Data capture was performed by Zodwa Zuma from Zokha Statistical Consultancy. Peter Alexander and Carin Runciman provided comment and edited the final report. Karien Brink from UJ Graphics designed the cover, and Postnet undertook the printing. 4

5 1. Introduction The 2016 Local Government Elections marked a new high point of electoral competition in South Africa. While retaining a majority of the vote nationwide, the ruling party, the African National Congress (ANC), conceded substantial ground to opposition parties, most notably the Democratic Alliance (DA) and the Economic Freedom Fighters (EFF). Crucially, the ANC lost its previously held majority in four of the country s major metropolitan municipalities: Nelson Mandela Bay in the Eastern Cape, and the City of Johannesburg, City of Tshwane, and Ekurhuleni in Gauteng. To better understand deepening electoral competition in South Africa, the Centre for Social Change at the University of Johannesburg conducted an exit poll of voters on 3 August 2016, the day of the Local Government Elections. The project included 110 fieldworkers conducting surveys in 11 different sites. A total of 4,313 voters participated in the survey. This report presents some preliminary and provisional findings from the 2016 Local Government Election Survey (LGES16). Further reading based on a similar survey conducted by the Centre for Social Change during the 2014 National Elections is provided at the end of the report. 2. Methodology The 2016 Local Government Election Survey (LGES16), conducted by the Centre for Social Change at the University of Johannesburg, took place in the 11 sites listed in Table 1. These sites should not be taken as nationally representative. Impoverished black townships and informal settlements are over-represented. According to the 2011 Census, for example, 77 percent of South African residents are Black/African, and 14 percent live in informal dwellings. Within the survey sample, however, 94 percent of respondents were Black/African and 34 percent of respondents lived in informal dwellings (shacks). Site selection was largely determined by places where previous research had been undertaken. Eight sites were surveyed during the 2014 National Election and one site (Balfour) was surveyed during the 2011 Local Government Election. Only two sites were new. These were Freedom Park, where a student affiliated to the Centre for Social Change is conducting research, and Hammankraal, chosen because we wanted to include a site in Tshwane. In most cases the sites are locations where members of the Centre for Social Change have conducted research or have ongoing projects. This selection strategy will enable us to situate our findings within a comparative and historical context in later publications. The goal was to cover selected voting stations for the entire day, from opening at 7:00am to closing at 7:00pm. To achieve this goal, two fieldwork teams were assigned to each site: a morning team to cover 7:00am-1:00pm, and an 5

6 afternoon team to cover 1:00pm to 7:00pm. The one exception was Freedom Park, which only had a single fieldwork team working throughout the day. Covering a single voting station proved to be difficult in practice. A few teams experienced intimidation from local political party activists, and in some instances the flow of voters was too slow to obtain a sufficient number of surveys. In one instance (Potchefstroom), the fieldwork team decided to split into two for strategic reasons, in order to cover both the main town and the nearby informal settlement. Due to these various factors, the fieldwork teams covered multiple voting stations in five of the eleven sites. In total the survey was administered in 21 different voting stations. Table 1. Survey sites: voting districts, wards, municipalities, province, and sample size Site name Voting districts Wards Municipality Province Sample size Marikana , , Potchefstroom , , , Balfour , Zamdela , , , , , , , , , Madibeng, Rustenberg Ventersdorp/ Tlokwe North West 588 North West 266 Dipaleseng Mpumalanga 309 Metsimaholo Free State 386 Hammanskraal Tshwane Gauteng 547 Freedom Park , Johannesburg Gauteng Thembelihle Johannesburg Gauteng 354 Motsoaledi Johannesburg Gauteng 382 Alexandra Johannesburg Gauteng 407 Brixton Johannesburg Gauteng 440 University of Johannesburg Johannesburg Gauteng 425 Source: South African Research Chair in Social Change LGES16. The survey was administered to voters directly outside of the voting stations, beyond the perimeter established by the Electoral Commission of South Africa (IEC). Respondents were recruited following the process of casting their vote. Across the eleven sites, the survey was administered to 4,313 voters. The specific sample sizes for each of the sites are provided in the far-right column of Table 1, above. The total survey sample accounts for 16 percent of the total number 6

7 of voters who cast votes at the 21 voting stations where the survey was administered. In other words, nearly one out of every six voters participated in the survey. As shown in Figure 1, this rate of participation varied considerably across sites. These figures represent the percentage of actual voters, across all of the voting stations for the given site, who participated in the survey. The participation rates ranged from 5-7 percent of voters in Freedom Park and Potchefstroom, to approximately half of voters in Brixton and the University of Johannesburg. The participation rate was lower in Freedom Park because there was a single research team that only covered part of the day, and the participation rate in Potchefstroom was lower because the research teams covered four different voting stations. From fieldworker reports it was clear that only a small minority or voters refused to complete the survey. Figure 1. Percentage of voters who participated in the survey, by site % 0 Source: South African Research Chair in Social Change LGES16; IEC. The survey itself included 21 items. The questions covered the following: age, gender, mother tongue (primary language), race, employment and student status, receipt of a social grant, smart phone ownership, level of satisfaction with democracy, support for a new workers political party, participation in various forms of protest activity, and who the individual voted for in both the 2014 (national) and 2016 (local) elections. A profile of respondents in the survey sample is provided in Appendix 1. 7

8 3. Voting patterns and comparison to actual voting results The survey asked respondents to indicate both the political party of the Ward Councilor candidate that they voted for, and the political party that they voted for on the Proportional Representation (PR) ballot. Among those respondents who were willing to indicate their voting decisions, 96 percent of respondents chose the same party on both ballots. This report focuses on how respondents voted on the PR ballot. Compared to the national results, ANC and EFF voters are over-represented in the survey sample while DA voters and voters for other political parties are underrepresented. Among those respondents who indicated their vote on the PR ballot, 57 percent voted for the ANC, compared to 54 percent in the actual results, and 23 percent voted for the EFF, compared to 8 percent in the actual results. Conversely, 13 percent of respondents voted for the DA, compared to 27 percent in the actual results, and 7 percent voted for other parties, compared to 11 percent in the actual results. Table 2. Actual results, comparison with survey results, and refusal rate, by site Vote percentages among actual votes cast Difference between survey % and actual % Refusal ANC EFF DA Other ANC EFF DA Other rate Marikana Potchefstroom Balfour Zamdela Hammanskraal Freedom Park Thembelihle Motsoaledi Alexandra Brixton University Total Source: South African Research Chair in Social Change LGES16; IEC. These differences were to be expected, given the over-representation of black voters and voters living in poor areas. More important, for our purposes, was obtaining a representative sample within each site. Table 2 provides, by site, the percentage of actual votes cast for the various political parties at the surveyed voting stations, and compares these results to the percentages among survey respondents. A positive result means that voters for the given party were oversampled, and a negative result means that voters for the given party were undersampled. In 8

9 Marikana, for example, ANC voters were undersampled by 12.3 percent, while EFF voters were oversampled by 11.5 percent. Table 2 also indicates the percentage of respondents who refused to reveal who they voted for. The result for the entire sample suggests that the survey is fairly representative of the voting stations covered, with the percentage of votes for each party among survey respondents falling within 4 percent of the actual results. But this overall result makes some unevenness between sites. ANC voters were oversampled by more than 10 percentage points in Potchefstroom, Alexandra, and Brixton, EFF voters were oversampled in Marikana, and voters for parties outside of the top three were oversampled in Freedom Park. One potential source of bias is the fact that one-quarter of the survey respondents refused to indicate the political party that they voted for. Many respondents maintained that their vote is my secret, and some may have been concerned about possible negative consequences if their voting decision was revealed. The presence of political party supporters outside of the voting stations fieldworkers noted that the ANC, in particular, had a strong presence at most of the voting stations may have amplified this effect. To the extent that supporters of certain parties were more likely to refuse to answer the voting question, this may partially explain the differences between the survey results and the actual voting results. The results below pertain to only those respondents who were willing to indicate the political party that they voted for on the PR ballot. 4. Voting patterns by age, gender, and ethnicity The survey revealed that voting patterns vary considerably by characteristics such as age, gender, and ethnicity. In general, older people, women, and isizulu speakers were more likely than other respondents to vote for the ANC. Figure 2 shows how voting patterns varied by age. Whereas roughly half of the respondents under 25 years-old voted for the ANC, more than 70 percent of respondents aged 55 and older did so. The EFF secured the greatest support among respondents between the ages of 25 and 39, while the DA did best among voters under the age of 25. Figure 3 presents voting patterns broken down by gender. Female respondents were substantially more likely than their male counterparts to vote for the ANC, and substantially less likely to vote for the EFF. Nearly two-thirds of women voted for the ANC, compared to only half of the men. Conversely, EFF secured votes from just under one-third of men, but only 14 percent of women. Within the survey sample, the DA secured roughly equal support from men and women. 9

10 Figure 2. Percentage of respondents voting for ANC, EFF, and DA, by age ANC EFF DA Source: South African Research Chair in Social Change LGES16 Figure 3. Percentage of respondents voting for ANC, EFF, and DA, by gender 70 66% % % % 14% 14% 0 Male Female ANC EFF DA Source: South African Research Chair in Social Change LGES16. 10

11 The survey included two questions related to ethnicity, one regarding the respondents mother tongue (primary language), and another pertaining to their racial identity. In practice the answers to these two questions are highly correlated. Figure 4 thus presents the voting patterns for specific combinations of language and race. The categories for isizulu, isixhosa, and Sepedi speakers pertain to respondents who also indicated Black/African racial identity. Due to small sample sizes, the Coloured/White category includes both English and Afrikaans speaking respondents. The results show that voting patterns vary substantially by ethnic identity. Matching the ethnic background of the ANC and EFF party leaders the ANC s Jacob Zuma is an isizulu speaker, the EFF s Julius Malema is a Sepedi speaker isizulu-speaking respondents were especially likely to vote for the ANC (69 percent), and Sepedi-speaking respondents were especially likely to vote for the EFF (31 percent). The far-right column of Figure 4 shows that Coloured and White voters voted overwhelmingly for the DA. The EFF also secured substantial support from isixhosa-speaking respondents (38 percent), but this result was largely driven by the Marikana site. Two-thirds of isixhosa-speaking respondents in Marikana voted for the EFF (67 percent), compared to only 11 percent in the other ten sites. This difference likely reflects the legacy of the Marikana massacre, during which the police killed 34 striking mineworkers and injured hundreds more. Of the 34 mineworkers that were killed, 26 were from the Eastern Cape, and it may be assumed that they were all isixhosa speakers. Figure 4. Percentage of respondents voting for ANC, EFF, and DA, by ethnicity 90 83% % 73% % 51% % 31% % 6% 5% 11% 7% 15% 7% 5% 0 isizulu isixhosa (all) isixhosa (non- Marikana) SePedi Coloured/White ANC EFF DA Source: South African Research Chair in Social Change LGES16. 11

12 5. Voting patterns by employment and student status Figure 5. Percentage of respondents voting for ANC, EFF, and DA, by employment and student status % 61% 63% 68% 52% % 12% 16% 15% 21% 14% 8% 8% 23% 19% 0 Full-time employed Full-time employed (non-marikana) Unemployed Unemployed (non- Marikana) Student ANC EFF DA Source: South African Research Chair in Social Change LGES16. The survey included five questions related to employment, student status, and participation in the labour force. One question asked about age, another asked whether the respondent was a student, and three questions asked about employment activities and desire for employment. For employment status, respondents were classified based on three questions with a skip sequence: 1) Do you have full-time employment? ; 2) If no, In the last month, did you do any casual, part-time, or temporary work for pay? ; 3) If no, Would you have liked to work last month? Those answering yes to #1, #2, and #3, respectively, are classified as full-time employed, precariously employed, and unemployed. Those answering no to all three questions, and students and those aged 65 or older who answered no to #1 and #2, are classified as not in the labour force. Figure 5 shows how voting patterns varied by employment and student status. The EFF was especially strong among full-time employed respondents, though this result was largely driven by the Marikana site. Whereas 29 percent of full-time employed respondents in the entire sample voted for the EFF, only 16 percent of those not in the Marikana sample did so. Conversely, the ANC secured a greater proportion of votes among unemployed respondents (63 percent) than they did among full-time employed respondents (52 percent). Unemployed respondents not in the Marikana site were especially likely to vote for the ANC (68 percent). Based on the survey results, the DA was especially strong among students. Whereas the DA only secured 13 percent of votes in the entire survey sample, 23 12

13 percent of the student respondents voted for the DA. Conversely, only 52 percent of students voted for the ANC, compared to 57 percent overall, and only 19 percent voted for the EFF, compared to 23 percent overall. 6. Voting patterns by indicators of government provision Figure 6. Percentage of respondents voting for ANC, EFF, and DA, by dwelling % % 59% 40 34% % 18% 19% 10 6% 5% 0 Shack RDP house Other brick/concrete ANC EFF DA Source: South African Research Chair in Social Change LGES16. The public provision of resources, such as social grants, housing, water, and electricity, is central to electoral competition in South Africa. Each of the major political parties addresses these issues in their campaigns, noting their capacity to deliver and where relevant, their performance record thus far. The survey provided a window into this dynamic with respect to two particular resources: housing and social grants. In both cases, evidence suggests that the ANC benefits greatly from the public provision of resources. Figure 6 indicates how voting patterns varied with respect to the type of dwelling in which respondents were living. More than three-quarters of respondents who were living in state-provided RDP housing voted for the ANC, compared to only half of those respondents who were living in shacks. 1 Conversely, shack dwellers were more than 2.5 times as likely as respondents living in RDP housing to vote for the EFF. The DA secured its greatest level of support from residents who 1 RDP housing refers to concrete or brick, single-family homes provided by the post-apartheid state. The acronym refers to the Reconstruction and Development Programme, which was put in place after the first democratic election in While the RDP was dispensed in 1996, state-provided homes are still popularly referred to as RDP houses. 13

14 were living in stand-alone brick or concrete housing that was not RDP housing, a hostel, residence hall, apartment or flat. Receipt of a social grant, such as a Child Support Grant, Disability Grant, or Older Persons Grant, also mattered for voting decisions, as illustrated in Figure 7. Social grant recipients were more likely than non-recipients to vote for the ANC, by 14 percentage points, and less likely to vote for the EFF (by 9 percentage points) and the DA (by 6 percentage points). Figure 7. Percentage of respondents voting for ANC, EFF, and DA, by social grant receipt % % % % 16% 8% 0 No social grant Social grant recipient ANC EFF DA Source: South African Research Chair in Social Change LGES Voting patterns by participation in protest activity Since the middle of the 2000s, and especially from 2009, South Africa has experienced growing popular resistance. This includes localised community protests around issues of service delivery, worker strikes, and more recently, a burgeoning student movement. The survey asked respondents whether or not they participated in three different kinds of protest action within the previous five years: a community protest or strike, such as a service delivery protest; a workers protest or strike; and a student protest or strike, such as the #FeesMustFall movement. The general pattern, as illustrated in Figure 8, was that protesters were more likely than non-protesters to vote for the EFF. Whereas only 16 percent of respondents who had not participated in any of the three kinds of protest action voted for the EFF, 29 percent of participants in community protests, 38 percent of participants in worker protests, and 24 percent of participants in student protests did so. Participants in student protests were also nearly twice as likely as participants in either community protests or worker protests to vote for the DA. 14

15 Figure 8. Percentage of respondents voting for ANC, EFF, and DA, by protest activity 70 64% 60 53% 52% 50 44% 40 38% 30 29% 24% % 9% 16% 16% 13% 0 Community protest Worker protest Student protest Had not participated in any protest ANC EFF DA Source: South African Research Chair in Social Change LGES Individual voter trajectories: from 2014 to 2016 The survey also provided a small window into the voting trajectories of individual voters over time. In addition to asking how they voted in the 2016 Local Government Election, respondents were also asked how they voted in the 2014 National Election. This allows for an analysis of the extent to which voters are staying loyal to certain political parties or changing their preference, and how these decisions are impacting specific political parties. Figure 9 compares how respondents voted in 2014 and 2016, focusing on four categories: ANC, DA, EFF, and other parties. Treating all other parties as equivalent, approximately two-thirds of respondents voted for the same political party in 2014 and The remaining one-third of respondents was split roughly equally between those who did not vote in the 2014 National Election (18 percent) and those who changed their political party of choice between 2014 and 2016 (14 percent). Figure 9. Comparing votes in 2014 (national) and 2016 (local) 15

16 Didn't vote % Changed vote 14% Same party 68% Same party Changed vote Didn't vote 2014 Source: South African Research Chair in Social Change LGES16. Figure 10. Percentage of respondents voting for ANC, EFF, and DA, by 2014 vote % 93% 83% % % 18% 8% 10% 5% 3% 5% 2% Didn't vote Voted ANC Voted EFF Voted DA ANC EFF DA Source: South African Research Chair in Social Change LGES16. 16

17 Figure 10 indicates how these patterns impacted specific political parties. Among those who did not vote in the 2014 National Election, just over half of respondents voted for the ANC in 2016 (53 percent), just under one-quarter of respondents voted for the EFF in 2016 (22 percent), and 18 percent of respondents voted for the DA in While three-quarters of new voters (though not necessarily first-time voters) voted for either the ANC or the EFF, this was slightly below the overall percentage in the entire sample (80 percent for ANC or the EFF). In contrast, the DA secured a higher percentage of votes among newly voting respondents (18 percent) than they did in the survey sample overall (13 percent). At a broader level, however, the comparison between voting decisions in 2014 and 2016 suggests that the EFF is leading the tide of change, even if that tide is relatively small. Overall, relatively few respondents switched political parties between 2014 and But EFF voters were especially loyal. Among respondents who voted for the EFF in 2014, 93 percent voted for the EFF again in The comparable rates for the ANC and DA were 10 percentage points lower, with only 83 percent voting for the respective party in Further, among those disaffected from the both the ANC and the DA, the EFF was the most popular choice in Among respondents who voted for the ANC in 2014, 8 percent voted for the EFF in 2016, compared to only 5 percent for the DA. Among respondents who voted for the DA in 2014, 9 percent voted for the EFF in 2016, compared to only 5 percent for the ANC. 9. Conclusion This report has provided some preliminary findings about the 2016 Local Government Election based on a survey of voters as they exited voting stations on 3 August. The survey was conducted at 21 voting stations in Gauteng, North West, Mpumalanga and Free State, most of them in townships and informal settlements. In total, 4,313 questionnaires were completed, with this amounting to nearly 16 percent of voters in surveyed locations. While the findings are not nationally representative, comparison with IEC data suggests they are broadly representative of the voting districts where the study was undertaken. Thus, they provide valuable data about working-class voters. We show that the ANC is especially strong among older voters and the EFF and DA are stronger among those who are younger. While the ANC has more support among female than male voters, the contrary is true for the EFF, and the DA s support is divided fairly evenly. Results from questions on housing and social benefits show that the ANC did particularly well among voters living in RDP houses and/or receiving a social benefit, and the EFF achieved its best results among shack dwellers. EFF voters were more likely to have participated in protest activity than ANC voters. Comparing this election with the 2014 National Election, 68 percent of survey respondents voted for the same party and 14 percent changed parties. The 17

18 most common change, accounting for one-third of those who did so, was from the ANC to the EFF. Future publications will examine the 2016 Local Government Election Survey in more detail. This will include the following: more sophisticated statistical analysis of voting trends; an analysis of the reasons that respondents provided for their voting decisions; a study of attitudes towards democracy and the possibility of a new political party for workers. Exit surveys provide valuable data for understanding elections, the most widely discussed activity in a democratic society. They supplement information obtained from election results, opinion polls and qualitative research, adding to informed debate. We recommend that in 2019 a nationally representative survey be conducted. 18

19 Appendix 1. A profile of survey respondents Selected characteristics of survey respondents Characteristic Percentage in sample Age Gender Male 51.1 Female 48.9 Ethnicity isizulu 23.9 isixhosa 18.2 SeTswana 18.2 SeSotho 17.3 Sepedi 6.1 XiTsonga 4.7 TshiVenda 2.1 SiSwati 1.6 isindebele 1.4 Black/African-other 0.7 Coloured/White 4.9 Other 1.0 Employment/student status Full-time 37.0 Precarious 25.7 Unemployed 15.3 Not in labour force 22.0 Student 24.1 Dwelling Informal (shack) 33.8 RDP house 17.5 Brick/concrete standalone 27.1 Hostel/Residence Hall 13.5 Other

20 Appendix 2. Further reading The following publications discuss findings from the 2011 Local Government Election, the 2014 National Election and the 2014 National Election Survey conducted by the Centre for Social Change at the University of Johannesburg. Peter Alexander, 2012, Barricades, Ballots and Experimentation: Making Sense of the 2011 Local Government Election with a Social Movement Lens, Pp in Contesting Transformation: Popular Resistance in Twenty-First Century South Africa, edited by Marcelle C. Dawson and Luke Sinwell. London: Pluto Press. Marcel Paret, 2016, Contested ANC Hegemony in the Urban Townships: Evidence from the 2014 South African Election African Affairs 115(460): Mahlatse Rampedi, 2015, Youth, Protest and Political Participation: The Case of Zamdela, Sasolburg, Unpublished MA dissertation, University of Johannesburg, Johannesburg. Carin Runciman, Forthcoming, The Ballot and the Brick : Protest, Voting and Non-Voting in Post-Apartheid South Africa, Journal of Contemporary African Studies. Anastasia Ryabchuk, Forthcoming, Voter abstention in South African 2014 elections: beyond the apathy argument, Transformation. 20

21 21